Representational learning by optimization of neural manifolds in an olfactory memory network

Area of Science:

• Neuroscience
• Computational Neuroscience
• Olfactory System

Background:

• Higher brain functions rely on experience-dependent neural representations.
• These representations may be organized by attractor dynamics or geometric changes in neural manifolds.

Purpose of the Study:

• Investigate neural representations in zebrafish (pDp) area during olfactory learning.
• Differentiate between attractor dynamics and geometric manifold modifications for information storage.

Main Methods:

• Analyzed odor-evoked activity in zebrafish (pDp) of juvenile and adult.
• Applied manifold capacity analysis to quantify representational geometry.
• Correlated manifold geometry with odor discrimination performance.

Main Results:

• No clear signatures of attractor dynamics were found in the (pDp) area.
• Olfactory training enhanced the separation of neural manifolds for task-relevant odors.
• Manifold capacity predicted individual odor discrimination ability.

Conclusions:

• The zebrafish (pDp) area stores information via the geometry of neural manifolds.
• This geometric organization supports sensory and semantic mapping for learning.
• Findings align with autoassociative network models and synaptic balance.